Preclinical and clinical studies showed that treatment with recombinant granulocyte colony stimulating factor (G-CSF) could improve host defense and reduce the risk of infection and sepsis in neutropenic patients. Preclinical studies suggested that treatment with G-CSF might also be beneficial during sepsis in the immunocompetent host. Despite these findings, G-CSF in early studies has not been beneficial in nonneutropenic patients with sepsis. Possibly consistent with this clinical experience, we found in a series of canine studies, that G-CSF treatment had variable effects with E. coli bacteria challenges at differing sites and in differing dosages. To better determine the influence of these two factors on the effects of G-CSF we developed a rat model of sepsis incorporating a multifactorial study design which permits simultaneous assessment of multiple variables. In these studies, with intravenous E. coli challenge, G-CSF increased lethality at all bacterial dosages. However, with intrabronchial and intraperitoneal challenge, G-CSF improved survival with low and high bacterial dosages, but appeared harmful with an intermediate dose. Laboratory investigations suggest that the variable effects of G-CSF on outcome in this study are related in part to its ability to increase circulating neutrophil numbers and the clearance of microbial and host inflammatory mediators. High levels of circulating bacteria were associated with inhibition of both of these G-CSF effects. Clinically, this study suggests that G-CSF may have its greatest beneficial effect in patients with extravascular infection and limited bac-teremia. Further studies are being designed to determine whether the harmful effects of G-CSF with intravascular challenge are related specifically to endothelial injury.